/// <exception cref="System.Exception"/>
public virtual void TestInvalidatedAMHostPortOnAMRestart()
{
MockRM rm1 = new MockRM(conf);
rm1.Start();
MockNM nm1 = new MockNM("127.0.0.1:1234", 15120, rm1.GetResourceTrackerService());
nm1.RegisterNode();
// a failed app
RMApp app2 = rm1.SubmitApp(200);
MockAM am2 = MockRM.LaunchAndRegisterAM(app2, rm1, nm1);
nm1.NodeHeartbeat(am2.GetApplicationAttemptId(), 1, ContainerState.Complete);
am2.WaitForState(RMAppAttemptState.Failed);
rm1.WaitForState(app2.GetApplicationId(), RMAppState.Accepted);
// before new attempt is launched, the app report returns the invalid AM
// host and port.
GetApplicationReportRequest request1 = GetApplicationReportRequest.NewInstance(app2
.GetApplicationId());
ApplicationReport report1 = rm1.GetClientRMService().GetApplicationReport(request1
).GetApplicationReport();
NUnit.Framework.Assert.AreEqual("N/A", report1.GetHost());
NUnit.Framework.Assert.AreEqual(-1, report1.GetRpcPort());
}
// This is to test AM Host and rpc port are invalidated after the am attempt
// is killed or failed, so that client doesn't get the wrong information.
/// <exception cref="System.Exception"/>
public virtual void TestInvalidateAMHostPortWhenAMFailedOrKilled()
{
conf.SetInt(YarnConfiguration.RmAmMaxAttempts, 1);
MockRM rm1 = new MockRM(conf);
rm1.Start();
// a succeeded app
RMApp app1 = rm1.SubmitApp(200);
MockNM nm1 = new MockNM("127.0.0.1:1234", 15120, rm1.GetResourceTrackerService());
nm1.RegisterNode();
MockAM am1 = MockRM.LaunchAndRegisterAM(app1, rm1, nm1);
MockRM.FinishAMAndVerifyAppState(app1, rm1, nm1, am1);
// a failed app
RMApp app2 = rm1.SubmitApp(200);
MockAM am2 = MockRM.LaunchAndRegisterAM(app2, rm1, nm1);
nm1.NodeHeartbeat(am2.GetApplicationAttemptId(), 1, ContainerState.Complete);
am2.WaitForState(RMAppAttemptState.Failed);
rm1.WaitForState(app2.GetApplicationId(), RMAppState.Failed);
// a killed app
RMApp app3 = rm1.SubmitApp(200);
MockAM am3 = MockRM.LaunchAndRegisterAM(app3, rm1, nm1);
rm1.KillApp(app3.GetApplicationId());
rm1.WaitForState(app3.GetApplicationId(), RMAppState.Killed);
rm1.WaitForState(am3.GetApplicationAttemptId(), RMAppAttemptState.Killed);
GetApplicationsRequest request1 = GetApplicationsRequest.NewInstance(EnumSet.Of(YarnApplicationState
.Finished, YarnApplicationState.Killed, YarnApplicationState.Failed));
GetApplicationsResponse response1 = rm1.GetClientRMService().GetApplications(request1
);
IList <ApplicationReport> appList1 = response1.GetApplicationList();
NUnit.Framework.Assert.AreEqual(3, appList1.Count);
foreach (ApplicationReport report in appList1)
{
// killed/failed apps host and rpc port are invalidated.
if (report.GetApplicationId().Equals(app2.GetApplicationId()) || report.GetApplicationId
().Equals(app3.GetApplicationId()))
{
NUnit.Framework.Assert.AreEqual("N/A", report.GetHost());
NUnit.Framework.Assert.AreEqual(-1, report.GetRpcPort());
}
// succeeded app's host and rpc port is not invalidated
if (report.GetApplicationId().Equals(app1.GetApplicationId()))
{
NUnit.Framework.Assert.IsFalse(report.GetHost().Equals("N/A"));
NUnit.Framework.Assert.IsTrue(report.GetRpcPort() != -1);
}
}
}
// Test Kill an app while the app is failing
/// <exception cref="System.Exception"/>
public virtual void TestKillFailingApp()
{
// this dispatcher ignores RMAppAttemptEventType.KILL event
Dispatcher dispatcher = new _AsyncDispatcher_708();
MockRM rm1 = new _MockRM_731(dispatcher, conf);
rm1.Start();
MockNM nm1 = new MockNM("127.0.0.1:1234", 8192, rm1.GetResourceTrackerService());
nm1.RegisterNode();
RMApp app1 = rm1.SubmitApp(200);
MockAM am1 = MockRM.LaunchAndRegisterAM(app1, rm1, nm1);
rm1.KillApp(app1.GetApplicationId());
// fail the app by sending container_finished event.
nm1.NodeHeartbeat(am1.GetApplicationAttemptId(), 1, ContainerState.Complete);
rm1.WaitForState(am1.GetApplicationAttemptId(), RMAppAttemptState.Failed);
// app is killed, not launching a new attempt
rm1.WaitForState(app1.GetApplicationId(), RMAppState.Killed);
}
// Test Kill an app while the app is finishing in the meanwhile.
/// <exception cref="System.Exception"/>
public virtual void TestKillFinishingApp()
{
// this dispatcher ignores RMAppAttemptEventType.KILL event
Dispatcher dispatcher = new _AsyncDispatcher_654();
MockRM rm1 = new _MockRM_677(dispatcher, conf);
rm1.Start();
MockNM nm1 = new MockNM("127.0.0.1:1234", 8192, rm1.GetResourceTrackerService());
nm1.RegisterNode();
RMApp app1 = rm1.SubmitApp(200);
MockAM am1 = MockRM.LaunchAndRegisterAM(app1, rm1, nm1);
rm1.KillApp(app1.GetApplicationId());
FinishApplicationMasterRequest req = FinishApplicationMasterRequest.NewInstance(FinalApplicationStatus
.Succeeded, string.Empty, string.Empty);
am1.UnregisterAppAttempt(req, true);
rm1.WaitForState(am1.GetApplicationAttemptId(), RMAppAttemptState.Finishing);
nm1.NodeHeartbeat(am1.GetApplicationAttemptId(), 1, ContainerState.Complete);
rm1.WaitForState(am1.GetApplicationAttemptId(), RMAppAttemptState.Finished);
rm1.WaitForState(app1.GetApplicationId(), RMAppState.Finished);
}
/// <exception cref="System.Exception"/>
private void AmRestartTests(bool keepRunningContainers)
{
MockRM rm = new MockRM(conf);
rm.Start();
RMApp app = rm.SubmitApp(200, "name", "user", new Dictionary <ApplicationAccessType
, string>(), false, "default", -1, null, "MAPREDUCE", false, keepRunningContainers
);
MockNM nm = new MockNM("127.0.0.1:1234", 10240, rm.GetResourceTrackerService());
nm.RegisterNode();
MockAM am0 = MockRM.LaunchAndRegisterAM(app, rm, nm);
int NumContainers = 1;
// allocate NUM_CONTAINERS containers
am0.Allocate("127.0.0.1", 1024, NumContainers, new AList <ContainerId>());
nm.NodeHeartbeat(true);
// wait for containers to be allocated.
IList <Container> containers = am0.Allocate(new AList <ResourceRequest>(), new AList
<ContainerId>()).GetAllocatedContainers();
while (containers.Count != NumContainers)
{
nm.NodeHeartbeat(true);
Sharpen.Collections.AddAll(containers, am0.Allocate(new AList <ResourceRequest>(),
new AList <ContainerId>()).GetAllocatedContainers());
Sharpen.Thread.Sleep(200);
}
// launch the 2nd container.
ContainerId containerId2 = ContainerId.NewContainerId(am0.GetApplicationAttemptId
(), 2);
nm.NodeHeartbeat(am0.GetApplicationAttemptId(), containerId2.GetContainerId(), ContainerState
.Running);
rm.WaitForState(nm, containerId2, RMContainerState.Running);
// Capture the containers here so the metrics can be calculated after the
// app has completed.
ICollection <RMContainer> rmContainers = rm.scheduler.GetSchedulerAppInfo(am0.GetApplicationAttemptId
()).GetLiveContainers();
// fail the first app attempt by sending CONTAINER_FINISHED event without
// registering.
ContainerId amContainerId = app.GetCurrentAppAttempt().GetMasterContainer().GetId
();
nm.NodeHeartbeat(am0.GetApplicationAttemptId(), amContainerId.GetContainerId(), ContainerState
.Complete);
am0.WaitForState(RMAppAttemptState.Failed);
long memorySeconds = 0;
long vcoreSeconds = 0;
// Calculate container usage metrics for first attempt.
if (keepRunningContainers)
{
// Only calculate the usage for the one container that has completed.
foreach (RMContainer c in rmContainers)
{
if (c.GetContainerId().Equals(amContainerId))
{
AggregateAppResourceUsage ru = CalculateContainerResourceMetrics(c);
memorySeconds += ru.GetMemorySeconds();
vcoreSeconds += ru.GetVcoreSeconds();
}
else
{
// The remaining container should be RUNNING.
NUnit.Framework.Assert.IsTrue("After first attempt failed, remaining container "
+ "should still be running. ", c.GetContainerState().Equals(ContainerState.Running
));
}
}
}
else
{
// If keepRunningContainers is false, all live containers should now
// be completed. Calculate the resource usage metrics for all of them.
foreach (RMContainer c in rmContainers)
{
AggregateAppResourceUsage ru = CalculateContainerResourceMetrics(c);
memorySeconds += ru.GetMemorySeconds();
vcoreSeconds += ru.GetVcoreSeconds();
}
}
// wait for app to start a new attempt.
rm.WaitForState(app.GetApplicationId(), RMAppState.Accepted);
// assert this is a new AM.
RMAppAttempt attempt2 = app.GetCurrentAppAttempt();
NUnit.Framework.Assert.IsFalse(attempt2.GetAppAttemptId().Equals(am0.GetApplicationAttemptId
()));
// launch the new AM
nm.NodeHeartbeat(true);
MockAM am1 = rm.SendAMLaunched(attempt2.GetAppAttemptId());
am1.RegisterAppAttempt();
// allocate NUM_CONTAINERS containers
am1.Allocate("127.0.0.1", 1024, NumContainers, new AList <ContainerId>());
nm.NodeHeartbeat(true);
// wait for containers to be allocated.
containers = am1.Allocate(new AList <ResourceRequest>(), new AList <ContainerId>())
.GetAllocatedContainers();
while (containers.Count != NumContainers)
{
nm.NodeHeartbeat(true);
Sharpen.Collections.AddAll(containers, am1.Allocate(new AList <ResourceRequest>(),
new AList <ContainerId>()).GetAllocatedContainers());
Sharpen.Thread.Sleep(200);
}
rm.WaitForState(app.GetApplicationId(), RMAppState.Running);
// Capture running containers for later use by metrics calculations.
rmContainers = rm.scheduler.GetSchedulerAppInfo(attempt2.GetAppAttemptId()).GetLiveContainers
();
// complete container by sending the container complete event which has
// earlier attempt's attemptId
amContainerId = app.GetCurrentAppAttempt().GetMasterContainer().GetId();
nm.NodeHeartbeat(am0.GetApplicationAttemptId(), amContainerId.GetContainerId(), ContainerState
.Complete);
MockRM.FinishAMAndVerifyAppState(app, rm, nm, am1);
// Calculate container usage metrics for second attempt.
foreach (RMContainer c_1 in rmContainers)
{
AggregateAppResourceUsage ru = CalculateContainerResourceMetrics(c_1);
memorySeconds += ru.GetMemorySeconds();
vcoreSeconds += ru.GetVcoreSeconds();
}
RMAppMetrics rmAppMetrics = app.GetRMAppMetrics();
NUnit.Framework.Assert.AreEqual("Unexcpected MemorySeconds value", memorySeconds,
rmAppMetrics.GetMemorySeconds());
NUnit.Framework.Assert.AreEqual("Unexpected VcoreSeconds value", vcoreSeconds, rmAppMetrics
.GetVcoreSeconds());
rm.Stop();
return;
}
// The test verifies processing of NMContainerStatuses which are sent during
// NM registration.
// 1. Start the cluster-RM,NM,Submit app with 1024MB,Launch & register AM
// 2. AM sends ResourceRequest for 1 container with memory 2048MB.
// 3. Verify for number of container allocated by RM
// 4. Verify Memory Usage by cluster, it should be 3072. AM memory + requested
// memory. 1024 + 2048=3072
// 5. Re-register NM by sending completed container status
// 6. Verify for Memory Used, it should be 1024
// 7. Send AM heatbeat to RM. Allocated response should contain completed
// container.
/// <exception cref="System.Exception"/>
public virtual void TestProcessingNMContainerStatusesOnNMRestart()
{
conf.SetInt(YarnConfiguration.RmAmMaxAttempts, 1);
MemoryRMStateStore memStore = new MemoryRMStateStore();
memStore.Init(conf);
// 1. Start the cluster-RM,NM,Submit app with 1024MB,Launch & register AM
MockRM rm1 = new MockRM(conf, memStore);
rm1.Start();
int nmMemory = 8192;
int amMemory = 1024;
int containerMemory = 2048;
MockNM nm1 = new MockNM("127.0.0.1:1234", nmMemory, rm1.GetResourceTrackerService
());
nm1.RegisterNode();
RMApp app0 = rm1.SubmitApp(amMemory);
MockAM am0 = MockRM.LaunchAndRegisterAM(app0, rm1, nm1);
// 2. AM sends ResourceRequest for 1 container with memory 2048MB.
int noOfContainers = 1;
IList <Container> allocateContainers = am0.AllocateAndWaitForContainers(noOfContainers
, containerMemory, nm1);
// 3. Verify for number of container allocated by RM
NUnit.Framework.Assert.AreEqual(noOfContainers, allocateContainers.Count);
Container container = allocateContainers[0];
nm1.NodeHeartbeat(am0.GetApplicationAttemptId(), 1, ContainerState.Running);
nm1.NodeHeartbeat(am0.GetApplicationAttemptId(), container.GetId().GetContainerId
(), ContainerState.Running);
rm1.WaitForState(app0.GetApplicationId(), RMAppState.Running);
// 4. Verify Memory Usage by cluster, it should be 3072. AM memory +
// requested memory. 1024 + 2048=3072
ResourceScheduler rs = rm1.GetRMContext().GetScheduler();
int allocatedMB = rs.GetRootQueueMetrics().GetAllocatedMB();
NUnit.Framework.Assert.AreEqual(amMemory + containerMemory, allocatedMB);
// 5. Re-register NM by sending completed container status
IList <NMContainerStatus> nMContainerStatusForApp = CreateNMContainerStatusForApp(
am0);
nm1.RegisterNode(nMContainerStatusForApp, Arrays.AsList(app0.GetApplicationId()));
WaitForClusterMemory(nm1, rs, amMemory);
// 6. Verify for Memory Used, it should be 1024
NUnit.Framework.Assert.AreEqual(amMemory, rs.GetRootQueueMetrics().GetAllocatedMB
());
// 7. Send AM heatbeat to RM. Allocated response should contain completed
// container
AllocateRequest req = AllocateRequest.NewInstance(0, 0F, new AList <ResourceRequest
>(), new AList <ContainerId>(), null);
AllocateResponse allocate = am0.Allocate(req);
IList <ContainerStatus> completedContainersStatuses = allocate.GetCompletedContainersStatuses
();
NUnit.Framework.Assert.AreEqual(noOfContainers, completedContainersStatuses.Count
);
// Application clean up should happen Cluster memory used is 0
nm1.NodeHeartbeat(am0.GetApplicationAttemptId(), 1, ContainerState.Complete);
WaitForClusterMemory(nm1, rs, 0);
rm1.Stop();
}
